3 May 2024
Dear HKUST SuperPOD users,
This is a follow-up of the recent adjustment of SuperPOD resources that was implemented early this week. We noticed that many normal projects had pending jobs after we reserved some nodes for large projects appointed by Office of Provost. As mentioned in our message, we are working on an enhanced setup for normal projects to utilize idle resources in these reserved nodes. We are pleased to give you a heads-up on this and provide further measures to optimize the utilization.
- We have introduced a new setting that allows users of normal projects to access idle resources of “large-project” partition. They can now submit jobs to “large-project” partition with same resources limitation as that of “project” partition, but their jobs might be pre-empted by higher priority jobs. These jobs can run for at least 2 hours before they are eligible for pre-emption. We recommend using checkpointing in your code to save your intermediate results before pre-emption may take place.
- The default wall time for “project” partition is changed to 8 hours with immediate effect. Maximum wall time remains intact which is 3 days. You are advised to specify the run time of the job allocation in your request. This can help for better scheduling and usually result in a shorter wait time for job admission.
For details of the new partition parameters, you can check https://itsc.hkust.edu.hk/services/academic-teaching-support/high-performance-computing/superpod/partitionandquota. We appreciate your patience and cooperation as we strive to optimize the system for all users. We will keep tabs on the resource utilization and make adjustment based on the latest situation as deemed necessary. Note that such changes may happen in short notice.
Regards,
Alan Wong
Head of Teaching Technologies
Information Technology Services Center
Hong Kong University of Science and Technology
Direct Line: (852) 2358-6246